The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection as it processes and acts upon the content of external files.
Ingestion points: Reads data from scripts/*.md and predictions/*.md and accepts user input to generate content for videos/<id>/script.md.
Boundary markers: The workflow does not specify the use of delimiters or clear instructions to ignore potential commands embedded within the processed markdown files.
Capability inventory: The skill is authorized to use Bash(*), Read, Write, and Edit tools, allowing for substantial file system modifications based on processed content.
Sanitization: There is no mention of sanitizing or validating the contents of the scripts or predictions before they are ingested into the agent context.

cheat-shoot